Fine-Tuning Platform Upgrades: Larger Models, Longer Contexts, Enhanced Hugging Face Integrations

Together AI · 10 months ago · 20

Together AI expanded its Fine-Tuning Platform to support training of large language models with over 100 billion parameters, including recent releases from DeepSeek, Qwen, and Meta. The platform now supports context lengths of up to 131,000 tokens for some models at no additional cost, and enables developers to fine-tune models from the Hugging Face Hub and save outputs directly back to it. These additions allow developers to customize state-of-the-art models for domain-specific tasks with lower inference costs and latency.